Specialized neurons in the right habenula mediate response to aversive olfactory cues

Hemispheric specializations are well studied at the functional level but less is known about the underlying neural mechanisms. We identified a small cluster of cholinergic neurons in the dorsal habenula (dHb) of zebrafish, defined by their expression of the lecithin retinol acyltransferase domain containing 2 a (lratd2a) gene and their efferent connections with a subregion of the ventral interpeduncular nucleus (vIPN). The lratd2a-expressing neurons in the right dHb are innervated by a subset of mitral cells from both the left and right olfactory bulb and are activated upon exposure to the odorant cadaverine that is repellent to adult zebrafish. Using an intersectional strategy to drive expression of the botulinum neurotoxin specifically in these neurons, we find that adults no longer show aversion to cadaverine. Mutants with left-isomerized dHb that lack these neurons are also less repelled by cadaverine and their behavioral response to alarm substance, a potent aversive cue, is diminished. However, mutants in which both dHb have right identity appear more reactive to alarm substance. The results implicate an asymmetric dHb-vIPN neural circuit in the processing of repulsive olfactory cues and in modulating the resultant behavioral response.


Introduction
Fish use the sense of smell to search for food, detect danger, navigate, and communicate social information by detecting chemical cues in their aquatic environment (Yoshihara, 2014). As with birds and mammals, perception of olfactory cues is lateralized and influences behavior (Siniscalchi, 2017). In zebrafish, nine glomerular clusters in the olfactory bulb (OB) receive olfactory information from sensory neurons in the olfactory epithelium and, in turn, transmit signals to four forebrain regions: the posterior zone of the dorsal telencephalon (Dp), the ventral nucleus of the ventral telencephalon (Vv), the posterior tuberculum (PT), and the dorsal habenular region (dHb) (Miyasaka et al., 2014;Yoshihara, 2014). In contrast to all other target regions, which are located on both sides of the forebrain, only the right nucleus of the dHb is innervated by mitral cells that emanate from medio-dorsal (mdG) and ventro-medial (vmG) glomerular clusters in both OBs (Miyasaka et al., 2014;Yoshihara, 2014). Moreover, calcium imaging experiments suggest that the right dHb shows a preferential response to odorants compared to the left dHb (Chen et al., 2019;Dreosti et al., 2014;Jetti et al., 2014; Results lratd2a-expressing neurons in the right dHb receive bilateral olfactory input A subset of medio-dorsal mitral cells that are labeled by Tg(lhx2a:gap-YFP) were previously shown to project their axons bilaterally through the telencephalon and terminate in the right dHb , in the vicinity of a small population of lratd2a-expressing neurons [(deCarvalho et al., 2013) and Figure 1A]. To characterize this subset of dHb neurons, we used CRISPR/Cas9mediated targeted integration (Kimura et al., 2014) to introduce the QF2 transcription factor (Ghosh and Halpern, 2016;Subedi et al., 2014) under the control of lratd2a transcriptional regulation ( Figure 1B). QF2 does not disrupt transcription at the lratd2a locus ( Figure 1-figure supplement 1) and drives expression of QUAS-regulated genes encoding fluorescent protein reporters in a similar pattern to endogenous lratd2a gene expression in the nervous system (deCarvalho et al., 2013). In 5 days post fertilization (dpf) larval zebrafish, this includes a subset of neurons in the OB, the bilateral vHb, and asymmetrically distributed neurons in the dHb. There are 2.5 times more cells in the right dHb than the left ( Figure 1C-E'' and Figure 1-figure supplement 2) and those on the left consistently show weaker lratd2a expression and less GFP labeling than those on the right ( Figure 1A and Aversive olfactory cues activate lratd2a neurons in the right dHb Several studies using transgenic expression of the genetically encoded calcium indicator GCaMP have demonstrated that dHb neurons are activated by olfactory cues in larval zebrafish (Jetti et al., 2014;Krishnan et al., 2014). We had previously examined the habenular response of adult zebrafish to cadaverine, a known aversive olfactory cue that is released from decaying fish (Hussain et al., 2013), and to chondroitin sulfate, a component of alarm substance (also known as Schreckstoff), which is released from the skin of injured fish (Mathuru et al., 2012). However, we did not have the necessary transgenic tools to specifically record the activity of the lratd2a cell population (deCarvalho et al., 2014).
To determine whether olfactants activate the lratd2a-positive neuronal cluster in the dHb, we measured calcium signaling in Tg(lratd2a:QF2) c644 , Tg(QUAS:GCaMP6f) c587 larvae at 7 dpf. We monitored the responses of individual cells in the right dHb, as GCaMP6f labeling was weakly or not detected in neurons on the left (data not shown). The observed changes in fluorescence intensity after the addition of two applications of an aversive cue were compared to those observed after two additions of vehicle alone (i.e. deionized water), all delivered 1 min apart (Figure 2A and B). Cadaverine elicited a significant increase in GCaMP6f labeling in the same neurons following the first and second applications (16-fold increase averaged over both doses; Figure 2C). On average, a greater than     (C-D) Change in the intensity of GCaMP6f fluorescence for lratd2a neurons in the right dHb in response to consecutive delivery of (C) water or cadaverine [n = 30 neurons in three larvae, 0.009 ± 0.039 ΔF/F for water, 0.148 ± 0.074 ΔF/F for cadaverine] and of (D) water or chondroitin sulfate [n = 43 neurons in four larvae, 0.039 ± 0.037 ΔF/F for water, 0.176 ± 0.069 ΔF/F for chondroitin sulfate] at seven dpf, respectively. Two-way ANOVA reveals a significant effect of time [F (1, 29) = 32.09, p < 0.0001], interaction [F (3, 87) = 3.797, p = 0.0131] but no effect of vehicle vs. odorants [F (3, 87)  (E-G) Colocalization of fos and lratd2a transcripts in transverse sections of adult olfactory bulbs (left panels) and habenulae (middle panels) detected by double labeling RNA in situ hybridization 30 min after addition of (E) water, (F) cadaverine, or (G) alarm substance to the test tank. (E'-G') Higher magnification images (corresponding to dashed boxes in E-G) show lratd2a (brown) coexpressed with fos (blue) in cells of the right dHb (arrowheads). Scale bars, 100 μm. (H) Quantification of fos-expressing cells in the adult dHb after addition of water [3.58 ± 0.811 cells in the left and 5.61 ± 0.85 in the right dHb, n = 31 sections from 16 adult brains], cadaverine [5.32 ± Figure 2 continued on next page fourfold increase in calcium signaling was measured in individual lratd2a-expressing neurons following addition of chondroitin sulfate compared to their response to vehicle alone ( Figure 2D).
To examine whether the response to aversive odorants persists in the olfactory-dHb pathway of adult zebrafish, we used expression of the fos gene as an indicator of neuronal activation (deCarvalho et al., 2013;Hong et al., 2013). Consistent with previous findings (Dieris et al., 2017), cadaverine broadly activated OB mitral neurons in the dorsal glomerulus (dG), dorso-lateral glomerulus (dlG), medio-anterior glomerulus (maG), medio-dorsal glomerulus (mdG), lateral glomerulus (lG). In addition, we observed a threefold increase in the number of fos-expressing neurons in the right dHb following exposure of adult zebrafish to cadaverine relative to delivery of water alone (15.73 ± 1.25 vs 5.61 ± 1.07 cells, Figure 2E and F). The position of the fos-expressing cells in the right dHb corresponded to that of the lratd2a-expressing neurons (11.00% ± 1.07 of lratd2a-expressing region; Figure 2F'), supporting that the lratd2a subpopulation responds to cadaverine in both larvae and adults.
Exposure to alarm substance prepared from adult zebrafish increased the number of fos-expressing cells in the lateral glomerulus (lG) and dlG of the OB as would be expected (Mathuru et al., 2012;Yoshihara, 2014), but also in the dorso-lateral region of the dHb ( Figure 2G and G'), where transcripts co-localized to lratd2a-expressing neurons. In contrast to cadaverine, alarm substance activated neurons equally in both the left and right dHb (20.72 ± 2.70 cells on the left and 20.31 ± 2.53 on the right; Figure 2H).

Synaptic inhibition of right dHb lratd2a neurons reduces aversive response to cadaverine
To confirm that lratd2a expressing dHb neurons play a role in processing of aversive olfactory cues, we inhibited synaptic transmission in these cells and tested adults for their reaction to cadaverine. The odorant was introduced at one end of a test tank and the time individuals spent within or outside of this region of the tank was measured. We used a preference index that is based on the position of an individual fish at a given time relative to the application site of the odorant (refer to Materials and methods).
The online version of this article includes the following source data for figure 2: Source data 1. Source data associated with Figure     BoTxBLC-GFP was selectively expressed in lratd2a/slc5a7a neurons of the right dHb ( Figure 3D) in individuals bearing the three transgenes Tg(lratd2a:QF2) c601 , Tg(slc5a7a:Cre) c662 , and Tg(QUAS:loxP-mCherry-loxP-BoTxBLC-GFP) c674 . Axons labeled by BoTxBLC-GFP terminated at the vIPN ( Figure 3E), in the same location as those observed in Tg(lratd2a:QF2), Tg(QUAS:mApple-CAAX) fish ( Figure 1G) suggesting that botulinum neurotoxin inhibits synaptic transmission within this restricted region of the vIPN. We confirmed that BoTx-BLC-GFP labeling persisted in the right dHb neurons throughout development, although some variability was observed in the number of

BoTxBLC-GFP-positive cells between individuals (Figure 3-figure supplement 3).
To determine whether the lratd2a neurons in the right dHb contributed to the aversive response to cadaverine, we monitored the behavior within and between groups of adults that had or did not have the BoTxBLC-GFP transgene. Fish lacking the transgene showed a significantly reduced preference for the side of the tank where cadaverine had been applied ( Figure 3G).
When the response of individual fish within each group was compared over time ( Figure 3-figure view of GFP labeling in only the right dHb after Cre-mediated recombination in a 5 dpf Tg(lratd2a:QF2), Tg(slc5a7a:Cre), Tg(QUAS:loxP-mCherry-loxP-GFP-CAAX) larva. Scale bar, 25 μm. (D) BoTxBLC-GFP-labeled cells (open arrowhead) in the right dHb in Tg(lratd2a:QF2), Tg(slc5a7a:Cre), Tg(QUAS:loxP-mCherry-loxP-BoTxBLC-GFP) 5 dpf, 37 dpf, and 4 mpf zebrafish. Upper images show mCherry-labeled lratd2a Hb neurons, middle images show the subset of right dHb neurons that switched to GFP expression, and the bottom row are merged images. Scale bar, 50 μm. (E) Transverse section of BoTxBLC-GFP labeled axonal endings of dHb neurons that express Cre and lratd2a in a subregion of the vIPN (bracket) in Tg(lratd2a:QF2), Tg(slc5a7a:Cre), Tg(QUAS:loxP-mCherry-loxP-BoTxBLC-GFP) 37 dpf juveniles. Scale bar, 50 μm. (F, G) Preferred tank location prior to and after cadaverine addition of adults genotyped for absence (Cre-) or presence (Cre+) of Tg(slc5a7a:Cre). (F) Representative 1 min traces for single Cre-(blue) and Cre+ (purple) adults recorded over 10 min prior to (min 0-5) and after (min 6-10) addition of cadaverine to one end of the test tank (open arrows). (G) Preference index for all adults for an average of 2 min before (white) and for each of 3 min after (gray) the addition of cadaverine. In Cre-fish, aversive behavior was significantly increased at 2 min (p = 0.0116) and 3 min (p = 0.0344), n = 15 fish for each group. In contrast, Cre+ fish, showed no significant difference in their preferred location over time. Dashed red lines in F and G denote midpoint of test tank. Two-way ANOVA reveals significant effects of time [F (3, 27) = 29, p < 0.0001], but no effect of group [F (1, 14)         https://elifesciences.org/articles/72345/figures#video1 supplement 1), adults with or without BoTxBLC-GFP initially avoided the side of the test tank where cadaverine had been introduced. However, aversion was sustained for 4 min in control fish, but not in those expressing BoTxBLC-GFP in lratd2a neurons. These findings, from statistical tests on individuals both within and between groups, suggest that lratd2a neurons in the right dHb are required for a prolonged aversive response to cadaverine.
Disruption of synaptic transmission in lratd2a-expressing Hb neurons alone did not alter the response of zebrafish to alarm substance, which typically triggers erratic, rapid swimming and bottom dwelling, followed by freezing behavior (Diaz-Verdugo et al., 2019;Jesuthasan and Mathuru, 2008). Similar to controls, both juveniles and adults expressing BoTxBLC-GFP under the control of Tg(lratd2a:QF2) c601 showed rapid swimming/darting behavior within 22-25 s after delivery of alarm substance, first doubling their speed of swimming ( Figure 3H-K and Figure 3-figure supplement  4), and then freezing for the duration of the 5 min recording period. Blocking the activity of lratd2a neurons in the right dHb is therefore insufficient to diminish the robust behavioral changes elicited by alarm substance ( Figure 3H-K and Figure 3-figure supplement 4).

Zebrafish mutants with habenular defects show altered responses to aversive cues
We examined the response to aversive odorants by tcf7l2 zf55 mutants that develop with symmetric left-isomerized dHb and lack the vHb, but are viable to adulthood (Hüsken et al., 2014;Muncan et al., 2007). In agreement with the transformation of dHb identity, projections from OB mitral cells do not terminate in the right dHb of homozygous mutants nor are lratd2a-expressing neurons or their efferents to the vIPN detected ( Figure 4A-D).
Following application of cadaverine, tcf7l2 zf55 homozygous adults failed to exhibit the characteristic aversive behavior of their wild-type siblings ( Figure 4E and F and Figure 4-figure supplement 1). Exposure to alarm substance also did not elicit a significant increase in swimming speed from baseline (1.13 ± 0.22 cm/s before and 1.89 ± 0.56 cm/s after) relative to WT siblings (2.84 ± 0.48 cm/s before and 4.88 ± 0.63 cm/s after, Figure 4G, I and J). Homozygous tcf7l2 zf55 mutants tended to swim more slowly ( Figure 4G) and spend more time in the top half of a novel test tank than wild-type adults, although the latter behavior was suppressed in the presence of alarm substance ( Figure 4H). tcf7l2 zf55 mutants are thus partly responsive to this fearful cue.
To further assess the role of lratd2a-expressing neurons in aversive olfactory processing, we looked at homozygous mutants of the brain-specific homeobox (bsx) gene, which develop right-isomerized dHb [ (Schredelseker and Driever, 2018) and Figure 5A] and are viable to adulthood (Schredelseker and Driever, 2018). As might be expected when both dHb have right identity, equivalent populations of lratd2a-expressing neurons were found on both sides of the brain ( Figure 5C). Instead of innervating only the right dHb as in controls, the axons of lhx2a:gap-YFP-labeled olfactory mitral cells terminated in the left and right dHb [  and Figure 5B], where the clusters of lratd2a neurons are situated (data not shown). Projections from the lratd2a dHb neurons coursed bilaterally through the left and right fasciculus retroflexus (FR) and innervated the same limited region of the ventral IPN ( Figure 5C).
To measure the reaction to cadaverine in bsx m1376/m1376 adults with bilaterally symmetric lratd2a neurons, we counted the number of cells expressing fos in the dHb and found an increase in the left nucleus compared to heterozygous siblings ( Figure 5D-E). Despite the symmetric activation of dHb neurons, bsx homozygotes and heterozygotes both showed reduced responsiveness to cadaverine ( Figure 5F and Figure 5-figure supplement 1). Overall, homozygous mutants were slower swimmers than heterozygotes ( Figure 5G-J); however, after exposure to alarm substance, their swimming speed relative to baseline was twofold faster than that of their heterozygous siblings ( Figure 5G), indicative of an enhanced response to this aversive cue.

Discussion
From worms to humans, stimuli including odors are differently perceived by left and right sensory organs to elicit distinct responses (Güntürkün and Ocklenburg, 2017;Güntürkün et al., 2020). Honeybees, for example, show an enhanced performance in olfactory learning when their right antenna is trained to odors (Guo et al., 2016;Letzkus et al., 2006;Rogers and Vallortigara, 2008).  In mice, over one third of mitral/tufted cells were found to be interconnected between the ipsilateral and contralateral olfactory bulbs for sharing of odor information received separately from each nostril, and for coordinated perception (Grobman et al., 2018). The zebrafish provides a notable example of a lateralized olfactory pathway, with the discovery of a subset of bilateral mitral cells that project to the dorsal habenulae but terminate only at the right nucleus (Miyasaka et al., 2014;Miyasaka et al., 2009). This finding prompted us to ask what is different about the post-synaptic dHb neurons that receive this olfactory input and what function does this asymmetric pathway serve.

Aversive olfactory cues activate identified neurons in the right dHb
We previously showed that the olfactory mitral cells that express lhx2a and are located in mediodorsal and ventro-medial bilateral glomerular clusters (Miyasaka et al., 2014;Miyasaka et al., 2009) and project their axons to a subregion of the right dHb where the lratd2a gene is transcribed (deCarvalho et al., 2013). From transgenic labeling with membrane-tagged fluorescent proteins and synaptophysin, we now confirm that the lhx2a olfactory neurons precisely terminate at a cluster of lratd2a/slc5a7a expressing cholinergic neurons present in the right dHb.
Through calcium imaging using a genetically encoded calcium indicator, we validated that the right dHb is responsive when larval zebrafish are exposed to aversive odors such as cadaverine or chondroitin sulfate (Jetti et al., 2014;Krishnan et al., 2014), a component of alarm substance (Mathuru et al., 2012), more specifically, that the lratd2a-expressing neurons of the right dHb significantly respond to these aversive olfactory cues above their response to water alone. As has also been observed by others (Jesuthasan et al., 2021), application of vehicle alone, even when introduced slowly into a testing chamber, is sufficient to elicit a change in GCaMP fluorescence.
In adults, we used fos expression as a measure of neuronal activation and showed that transcripts colocalized to lratd2a-expressing cells. Interestingly, cadaverine predominantly activated neurons in the right dHb in larvae and adults, whereas neurons responsive to alarm substance were detected in both the left and right dHb nuclei of adult zebrafish. Different types of olfactory cues activate distinct glomeruli in the OB (Friedrich and Korsching, 1997;Yoshihara, 2014), and consistent with the prior studies, we observed that, in adults, cadaverine significantly increased fos expression in the mdG and dG regions, the location of lhx2a neurons that project to the right dHb. By contrast, alarm substance predominantly activated neurons in the lG and dlG regions of the OB that innervate the telencephalon and posterior tuberculum (Miyasaka et al., 2014;Miyasaka et al., 2009), suggesting that both dHb receive input via this route rather than through direct olfactory connections. Indeed, we found that more neurons reacted to alarm substance than cadaverine throughout the brain, including in the Dp, Vv and thalamic areas (data not shown).
In previous experiments (deCarvalho et al., 2013), we did not detect activated neurons in the right dHb of adult zebrafish following exposure to cadaverine or alarm substance. Several factors could and WT sibling adults after application of cadaverine. (F) Preference index for mutants and WT siblings for an average of 2 min before (white) and for each of 3 min after (gray) the addition of cadaverine. Only WT fish showed a significant difference in their preferred location at 1 min (p = 0.0439) and at 2 min (p = 0.0184). For each group, n = 15 adults. Two-way ANOVA reveals a significant effect of time [F (3, 24) = 3.665, p = 0.046], group [F (1, 14) = 6.197, p = 0.026] and interaction [F (3, 30) = 7.953, p = 0.001]. Post-hoc analysis by Bonferroni's multiple comparisons. Dashed red lines denote midpoint of test tank. (G) Swimming speed for 30 s before and after addition of alarm substance was1.13 ± 0.22 cm/s and 1.89 ± 0.56 cm/s for tcf7l2 homozygotes and 2.84 ± 0.48 cm/s and 4.88 ± 0.63 cm/s for their WT siblings, n = 10 fish for each group. Two-way ANOVA reveals significant effects of time [F (1, 9) = 19.31, p = 0.0021] and group [F (1, 9) = 13.91, p = 0.0047], but no effect of interaction [F (1, 9)     Swimming speed for 30 s before and after addition of alarm substance. In heterozygous adults, swimming speed was 1.22 ± 0.31 cm/s before and 3.52 ± 0.44 cm/sec after and, in homozygotes, 0.46 ± 0.08 cm/s before and 2.80 ± 0.37 cm/s after, n = 15 adults Figure 5 continued on next page account for the difference from the earlier study: we now have the transgenic tools to examine lratd2a neurons directly, we used higher concentrations of cadaverine and alarm substance and, in contrast to delivering odorants to groups of zebrafish, we tested the neuronal response in individual adults.
It has been suggested that lateralized olfactory and visual functions of the dHb are more prominent early in development and less so at later stages (Fore et al., 2020). However, the presence of lratd2aexpressing neurons in the right dHb and their preferential response to cadaverine from larval to adult stages supports the persistence of lateralized activity and illustrates the value of examining defined neuronal populations.

Right dHb neurons mediate aversive behavioral responses
As a group, zebrafish in which the synaptic activity of lratd2a neurons in the right dHb was inhibited by BoTxBLC-GFP did not exhibit repulsion to cadaverine, although some individuals showed a mild aversive response that was not sustained relative to sibling controls. This subtle behavioral effect suggests that the entire population of lratd2a neurons may not be effectively inactivated in all animals. The described intersectional approach to suppress neuronal activity may be incomplete, as evidenced by the observed variability in BoTxBLC-GFP labeling between individual larvae. Additionally, compensatory mechanisms occurring between the onset of Cre expression in slc5a7a neurons (at 3 dpf) and adulthood could allow for partial recovery of the response to repulsive cues. It is also possible that the lratd2a neurons are a heterogeneous population, with different subsets controlling the magnitude of the aversive response or the duration. More finely tuned techniques for temporal or spatial regulation of neuronal inactivation would help resolve these issues.
Despite both being aversive cues (Hussain et al., 2013;Mathuru et al., 2012), cadaverine and alarm substance elicit different behavioral responses by adult zebrafish. Control fish show active repulsion to cadaverine for the first to 2-4 min of a 5 min testing period, whereas alarm substance triggers immediate erratic behavior such as rapid swimming and darting that is typically followed by prolonged freezing (Hussain et al., 2013;Mathuru et al., 2012). Therefore, it is not necessarily expected that the same neuronal populations will mediate the response to both substances.
Perturbation of the lratd2a-expressing right dHb neurons either selectively by BoTxBLC-mediated synaptic inactivation, or in tcf7l2 zf55 homozygous mutants that completely lack them, reduced aversion to cadaverine, either in the length or degree of the response. In contrast to juveniles or adults with BoTxBLC inactivated neurons that displayed a similar response to alarm substance as controls, tcf7l2 zf55 mutants, showed no difference in their swimming behavior before and after its addition. One explanation is that many regions throughout the brain are likely involved in directing the complex repertoire of behaviors elicited by alarm substance and inactivation of lratd2a neurons in the habenular region alone is insufficient to weaken the overall response. Furthermore, the tcf7l2 zf55 mutation could disrupt other brain regions that regulate behaviors elicited by alarm substance since the tfc7l2 gene is expressed in neurons throughout the brain, including the anterior tectum, dorsal thalamus and the hindbrain (Young et al., 2002).
Similar to tcf7l2 zf55 , the bsx m1376 mutation is pleiotropic resulting in right-isomerization of the dHb due to the absence of the parapineal (Schredelseker and Driever, 2018), and also the loss of the terminal tuberal hypothalamus, mammillary hypothalamic regions and secondary prosencephalon for each group. Two-way ANOVA reveals significant effect of time [F (1, 14) = 113.4, p < 0.0001], but no effect of group [F (1, 14)  The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Source data associated with Figure 5E-J.    (Schredelseker et al., 2020). We did not observe enhanced or prolonged aversion to cadaverine in bsx m1376 homozygotes relative to controls. However, although homozygous mutants did show a hyperactive response to alarm substance, we cannot discount the involvement of other affected brain regions. Albeit technically challenging in adults, a more selective test such as optogenetic activation of only the lratd2a dHb neurons in wild-type and mutant zebrafish could help resolve their contribution to the alarm response.
The identification of a subset of neurons in the right dHb that receive olfactory input and terminate their axons at a defined subregion of the ventral IPN lays the groundwork for tracing an entire pathway from olfactory receptors to the neurons directing the appropriate behavioral response. The midline IPN has been morphologically defined into subregions (deCarvalho et al., 2014 ;Lima et al., 2017;Quina et al., 2017), but their connectivity and functional properties are not well studied. Recent work has begun to assign different functions to given subregions, such as the role of the rostral IPN in nicotine aversion (Morton et al., 2018;Quina et al., 2017). Neurons in the ventral IPN project to the raphe nucleus (Agetsuma et al., 2010;Lima et al., 2017), but the precise identity of raphe neurons that are innervated by the lratd2a-expressing dHb neurons remains to be determined. Transcriptional profiling of the IPN should yield useful information on its diverse neuronal populations and likely lead to the identification of the relevant post-synaptic targets in the ventral IPN and their efferent connections. Elaboration of this pathway may also help explain the advantage of lateralization in the processing of aversive information. It has been argued, for instance, that the antennal specialization to aversive odors in bees is correlated with directed turning away from the stimulus and escape (Rogers and Vallortigara, 2019). Directional turning has also been observed in larval zebrafish (Horstick et al., 2020), but whether it is correlated with laterality of the Hb-IPN pathway is unclear. Beyond olfaction, left-right asymmetry appears to be a more general feature of stress-inducing, aversive responses as demonstrated for the rat ventral hippocampus (Sakaguchi and Sakurai, 2017) and human pre-frontal cortex, where heightened anxiety also activates more neurons on the right than on the left (Avram et al., 2010;Ocklenburg et al., 2016).

Continued on next page
Key resources
To produce Tol2 transposase mRNA, pCS-zT2TP was digested by NotI and RNA synthesized using the mMESSAGE mMACHINE SP6 Transcription Kit (AM1340, Thermo Fisher Scientific). RNA was purified by phenol/chloroform-isoamyl extraction, followed by chloroform extraction and isopropanol precipitation (Suster et al., 2011). A solution containing QF2/QUAS plasmid DNA (~25 ng/ μl), transposase mRNA (~25 ng/μl) and phenol red (0.5%) was microinjected into one-cell stage zebrafish embryos, which were raised to adulthood. To identify transgenic founders, F 0 adult fish were outcrossed to AB and embryos were assessed for the presence of the secondary marker by screening for mCherry labeling of hatching gland cells after 24 hr post fertilization (hpf) and raised to adulthood.

Preparation of odorants
Alarm substance was freshly prepared on the day of testing. Adult zebrafish (6 female and six male) were anesthetized in 0.02% tricaine (E10521, Ethyl 3-aminobenzoate methanesulfonate; Sigma-Aldrich). Shallow lesions were made on the skin (10 on each side) using a fresh razor blade and single fish were consecutively immersed in a beaker containing distilled water (25 ml for fos experiment and 50 ml for behavioral analyses) for 30 s at 4 °C. The solution was filtered using a 0.2 μm filter (565-0020, ThermoFisher Scientific) and stored at 4 °C until used (Mathuru et al., 2012). The cadaverine (#33211, final concentration 100 μM) and chondroitin sulfate (#c4384, 100 μg/ml) were purchased from Sigma-Aldrich, and stock solutions prepared in distilled water.

Calcium imaging in larval zebrafish
To monitor activity within the same population of lratd2a neurons in response to odorants or to a vehicle control, calcium signaling was performed on seven dpf Tg(lratd2a:QF2) c644 ; Tg(QUAS:G-CaMP6f) c587 larvae. Larvae were paralyzed by immersion in α-bungarotoxin (20 µl of 1 mg/ml solution in system water, B1601, ThermoFisher Scientific) Duboué et al., 2017;Severi et al., 2014 followed by washing in fresh system water. Individuals were embedded in 1.5% low melting agarose (SeaPlaque Agarose, Lonza) in a petri dish (60 mm) with a custom-designed mold. After solidification, the agarose around the nostrils was carefully removed with forceps for access to odorants, and the individual immersed in fresh system water. The dish was placed under a 20 X (NA = 0.5) water immersion objective on a Zeiss LSM 980 confocal microscope. Images were acquired in XYT acquisition mode with 512 × 200 pixel resolution at a rate of 2 Hz and digitized eight bit from single focal planes. The vehicle control (deionized water) followed by the odorant solution (0.2 ml) was slowly expelled through plastic tubing (Tygon R-3606; 0.8 mm ID, 2.4 mm OD) attached to a 1 ml syringe (BD 309659) and with the other end positioned in front of an individual's face. Two sequential applications of each were spaced by 1 min intervals. To calculate fluorescence intensity, regions of interest (ROI) were manually drawn around each cell in the average focal plane with the polygon tool and ROI manage in Fiji (Schindelin et al., 2012). To normalize calcium activity for each neuron to baseline fluorescence, the fractional change in fluorescence (ΔF/F) was calculated before the application of odorants (average of 320 frames from each neuron), according to the formula F = (F i -F 0 )/F 0 , where F i is the fluorescence intensity at a single time point and F 0 is the baseline fluorescence. All data and images were analyzed using custom programs in MATLAB (MathWorks, version 7.3), Excel software and GraphPad Prism9.

Assay of Fos expression in adult zebrafish
Individual adult zebrafish (7-9 months old) were placed in a tank with 1 L system water and acclimated for at least 1 hr prior to odorant exposure. Each odorant solution (1 ml) was gently pipetted into the tank water and the fish was kept there as the odorant diffused. After 30 min, the fish was sacrificed in an ice water slurry, and the brain dissected out and fixed in 4% paraformaldehyde in 1 X PBS overnight at 4 °C. Fixed brains were embedded in 4% low melting agarose in 1 X PBS and sectioned at 50 μm (for juvenile brains) or 70 μm (for adult brains) using a vibratome (VT1000S, Leica Biosystems, Inc). For more precise counting of fos expressing cells in adult brains, habenular sections were 35 μm thick. Sections were covered in 50% glycerol in 1 X PBS under coverslips. Bright-field images were captured with a Zeiss AxioCam HRc camera mounted on a Zeiss Axioskop. A Leica SP5 confocal microscope was used for fluorescent images. Data from fos RNA in situ hybridization experiments were quantified using ImageJ/Fiji software (Schindelin et al., 2012).

Behavioral assays
Behavioral assays were performed using 5-7 week-old juvenile zebrafish and adults that were of 4-8 months of age. Responses to odorants were measured between 10:00 a.m. and 4:00 p.m. and fish were starved for 1 day prior to testing (Koide et al., 2009). Individual adults were placed in a 1.5 L test tank (Aquatic Habitats) in 1 L of system water and allowed to acclimate for at least 1 hr. For experiments with juveniles, individuals were acclimated to the behavior room for 1 hr, gently netted into the test tank (20 × 9 x 8.3 cm, 1.5 L mating cage) containing 0.6 L system water and maintained there for 5 min prior to testing. Swimming activity was recorded for 5 min (4 min for juveniles) before and after the application of odorants. Odorants (2 ml for adults, 1 ml for juveniles) were slowly expelled through plastic tubing (Tygon R-3606; 0.8 mm ID, 2.4 mm OD) attached on one end to a 3 ml syringe (BD 309657) and on the other positioned at one end of the test tank. A preference index was calculated using the formula: Preference to odorant = [(Total time spent in the half of the tank where odorant was delivered) − (Total time spent in the other half of the tank)]/Total time (Koide et al., 2009;Wakisaka et al., 2017). A preference index of -1 indicates that the position of an individual fish is maximally distanced from the application site of the odorant.

Quantification and statistical analyses
Data were collected using custom written scripts in MATLAB (MathWorks). Statistical analyses were performed using GraphPad Prism9 (GraphPad Software, Inc). The Wilcoxon signed-rank test was used to compare a set of matched samples. An unpaired t-test was used to compare two independent data sets and two-way ANOVA followed by Bonferroni's post hoc test was used to compare datasets greater than two. All tests were two-tailed and results depicted as non-significant (ns, p> 0.05) or significant (*, p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001).

Additional files
Supplementary files • Supplementary file 1. Primers for making plasmids using the Gateway cloning system.
• Transparent reporting form • Source code 1. Example code file to run the MATLAB script.
• Source code 2. MATLAB script used for behavioral analyses and for running the example code file (Source Code File 1).

Data availability
All data generated or analyzed during this study are included in the manuscript and supporting file. Source data files have been provided for Figures 1-5. Behavioral analyses were performed using custom written scripts in MATLAB and uploaded as Source code Files.